> Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> > I got an interesting result. If I compile backend with -g (and without
> > -O2), I get no stuck spin lock errors. However, if s_lock.c is
> > compiled with -O2 enabled, I got the error again. It seems only
> > s_lock.c is related to this phenomenon.
>
> That's very interesting. Could optimization be breaking the TAS
> sequence on your platform? What is your platform, anyway?
> Might need to burrow into the assembly code to see just what's
> happening.
As I said, it's a x86 Linux (more precisely, kernel 2.2.18 with 2
processors, egcs 2.91). I suspect that the inlined TAS code might be
incompatible with the caller, s_lock.
--
Tatsuo Ishii